Liquidation heatmaps explained: what they show and how to read one
A liquidation heatmap estimates the prices at which leveraged positions would be force-closed by the exchange if the market got there. It is neither a record of past liquidations nor a forecast of where price is heading. It is an approximate snapshot of open positions, rebuilt from public data. This guide covers what the map shows, how it is computed and how to read it without asking more of it than it can give.
What is a liquidation heatmap?
On crypto perpetual futures, a trader who opens a leveraged position only posts part of its value as margin. If price moves far enough against them, that margin no longer covers the minimum the exchange requires, and the position is liquidated: the exchange's engine closes it at market. The level where this happens is the liquidation price.
A liquidation map applies that idea to the whole market. For every price level it estimates how much notional, in dollars, would be liquidated if the mark price reached it, split by side (longs and shorts) and by leverage. It is usually drawn as a heatmap over the candlestick chart: horizontal bands whose intensity shows how much money is at risk at each height.
The key word is "estimates". Centralised exchanges do not publish their users' positions or liquidation prices. Hyperliquid is the exception, because its positions are visible on chain. Everywhere else the map is an inference, and it should be read as one.
How a liquidation map is estimated from public data
Every perpetual contract has a buyer and a seller, and open interest counts how many contracts are still open. When open interest rises during a minute, new positions were opened in that minute, and their entry price is very close to the average traded price. When it falls, positions were closed. Following that trail minute by minute, the model rebuilds where leveraged money sits.
- Open interest: how many positions open or close at each step.
- 1-minute VWAP: the volume-weighted average price, used as the entry price.
- Mark price: the reference price exchanges liquidate against; it decides which levels have been crossed.
- Leverage distribution: how new notional splits across leverage levels, measured on real positions.
- Taker buy and sell volume: the side hitting the book with market orders tends to carry more leverage, so it gets more weight.
From entry price and leverage, the model computes where each slice liquidates, using the exchange's exact formula and maintenance margin. On liqlin the short timeframes (5m, 15m and 1h) update every minute and the long ones (4h and 1d) every 15 minutes. There is one map per exchange (Binance, Bybit, OKX and Hyperliquid) plus a combined view that adds the four without mixing their states.
Liquidation levels: longs below price, shorts above
A long profits when price rises, so it gets liquidated when price falls. A short is the mirror image. That is why long liquidations sit below the current price on the map and short liquidations sit above it. The higher the leverage, the thinner the margin behind the position and the closer its liquidation price is to the entry.
The table below works out the levels for an entry at 100,000 USDT, using the 0.4% maintenance margin rate of Binance's first BTC bracket:
| Leverage | Long liquidation | Distance | Short liquidation | Distance |
|---|---|---|---|---|
| 10x | 90,361.45 | −9.64% | 109,561.75 | +9.56% |
| 25x | 96,385.54 | −3.61% | 103,585.66 | +3.59% |
| 50x | 98,393.57 | −1.61% | 101,593.63 | +1.59% |
| 100x | 99,397.59 | −0.60% | 100,597.61 | +0.60% |
A 10x long opened at 100,000 liquidates at 90,361.45. At 100x it liquidates at 99,397.59, just 0.6% away. A 100x short from the same entry liquidates at 100,597.61. This is why high-leverage clusters hug the price while low-leverage ones sit much further out.
Reading the colours of a liquidation heatmap
Each cell is a narrow price interval at a given moment: 0.05% wide on short timeframes and 0.20% on long ones. Its colour shows how much estimated notional sits there compared with the rest of the map. Blue and green are thin levels; yellow and red are the heaviest.
On liqlin the colour reflects a cell's rank among all cells of the latest map, not its dollar amount. That keeps clusters visible even when the whole range holds money. A red band means the level is among the most loaded on that map, not that it exceeds a fixed figure. To compare two coins, look at the dollar amounts rather than the colours.
There are two ways to view it. "Recent", the default, paints only what was opened inside the visible window: each level starts at the candle where it was opened and runs until price crosses it. "All" paints everything still open, including positions from months ago, which is why it looks like bands running across the whole chart.
What happens when price reaches a liquidation level
When the mark price moves through a level, the positions estimated there are treated as liquidated or stopped out and removed from the map. That is why a band ends at the exact candle that touched it. If open interest drops without price crossing anything, the model shrinks every level by the same ratio, assuming closed positions are spread like the open ones.
The map does not say price will travel to a cluster. It says that if price gets there, a lot of notional is at risk at that level, and those liquidations are market orders that push in the direction of the move. What to do with that is each reader's own call; this guide is not investment advice.
Limits of a liquidation map, and how its accuracy is measured
Some things are invisible to the model. Cross-margin positions backed by a large balance liquidate much further away than their nominal leverage suggests. Positions hedged across exchanges add to open interest but never liquidate on price. Stop orders are not liquidations, although they often sit at similar levels. And the leverage mix comes from real Hyperliquid positions, the only public source that exposes them.
That is why liqlin calibrates the map against observed liquidations and measures its hit rate, pair by pair. The headline number answers a precise question: of the times price crossed a zone the map ranked in its top 5% by load, how often were there real liquidations? It is computed out of sample, on data the model was not fitted on.
Frequently asked questions
Does a liquidation heatmap predict where price will go?
No. It shows where positions are estimated to be at risk, not whether or when price will get there. Price can ignore a cluster for weeks or sweep it in minutes.
Why does the map use mark price instead of last traded price?
Exchanges liquidate against mark price, a reference price that stops a single isolated wick from triggering liquidations. Entry prices, on the other hand, come from the VWAP of trades, because positions are opened at the execution price.
Is the Binance map the same as the Hyperliquid map?
No. Each exchange has its own positions and its own map. On Hyperliquid positions are also public, and liqlin uses them to check the estimated map against the liquidation prices the exchange itself computes.
How much does the liqlin liquidation map cost?
The Hyperliquid BTC pair is free with a Google account. The Pro plan, with all four exchanges, the combined view and every other pair, costs $20 for 30 days.
Related maps
Guides to read the map
- The exact liquidation price formula, derived and explainedThe exact liquidation price formula for Binance USDⓈ-M perpetuals, derived step by step: MMR, margin brackets, cum and why the shortcut puts levels wrong.
- Where each leverage level gets liquidated, from 3x to 100xHow far from entry a 10x, 25x, 50x or 100x position gets liquidated, how a liquidation heatmap groups each leverage level and what changes in cross margin.
- Gold (XAUUSD), silver and oil liquidation mapWhere leveraged money gets liquidated in gold, silver, oil, EURUSD and USDJPY perpetuals on Binance, Bybit, OKX and Hyperliquid, shown with MT4 names.
- Accuracy: how we measure whether the map is rightHow we check whether a liquidation map is right: observed liquidations, hit rate by percentile, out-of-sample walk-forward tests and real Hyperliquid positions.